Teaching an Entire Generation AI
An Analysis of a Chinese AI General Education Textbook Series
Amidst a global focus on integrating artificial intelligence into general education, this analysis provides a systematic overview of a comprehensive AI textbook series from China's East China Normal University Press. The series is designed to create an integrated general education pathway that spans from primary and secondary school through to undergraduate studies. In 2024, more than 16 million children enrolled in primary schools across China, with a total enrolment of over 100 million students.
A core pedagogical philosophy underpins the series, emphasizing the use of contextual, real-world scenarios and narrative framing to lower the cognitive barrier to complex topics. This approach is augmented by the integration of local cultural elements to enhance content relevancy and a persistent inclusion of social and ethical considerations as a component of the learning process.
The curriculum is structured into three main tiers across seven volumes: a foundational "Getting Started" series for primary and secondary school, an "Advanced" series for later secondary and undergraduate students, and a capstone "Comprehensive Applications" volume for the high school level and above. The following sections detail the content of each series.
The "Getting Started" Series – Building AI Literacy and Interest
The initial "Getting Started" tier is positioned for primary and secondary school students, with the objective of demystifying AI by establishing foundational concepts and fostering interest through interactive experiences.
The content begins by introducing the concept of AI through its applications in daily life and proceeds to outline its developmental history, including foundational ideas such as the Turing test. Following this introduction, the series uses the autonomous vehicle as a central case study to systematically deconstruct five core concepts of AI: human-computer interaction, perception, machine learning, representation and reasoning, and social impact.
Building on this framework, subsequent chapters explore key application domains, explaining the fundamental principles of computer vision (e.g., the composition of images from pixels), natural language processing (e.g., word segmentation, sentiment analysis), and speech technology (e.g., the digitization of sound).
To make these concepts more concrete, the textbooks introduce specific technologies such as face landmark detection in facial recognition and keypoint detection in gesture recognition, allowing students to observe the practical details of AI applications.
The series also establishes two practical conceptual frameworks for students. The first is a "five-step AI workflow" (input, preprocess, feature extraction, match, output) that is applied across diverse applications like image classification and speech recognition. The second is an introduction to "prompt strategies" for interacting with large models, covering techniques such as setting clear goals and providing context.
The didactic method at this stage prioritizes experiential learning. It employs gamified and narrative-based activities, such as a "You Draw, I Guess" game or a couplet-matching exercise based on traditional Chinese literature, and utilizes a graphical programming platform to allow students to engage in AI model training without prior coding knowledge.
Ethical education is integrated from the outset. Modules include structured debates on the attribution of responsibility for autonomous vehicle accidents and discussions on the risks of deepfakes, intended to cultivate critical thinking about the societal implications of technology.
The "Advanced" Series – Deconstructing Core AI Technologies
The "Advanced" tier is designed to transition learners from a conceptual understanding to a technical one, targeting students from the later stages of secondary education through to the undergraduate level. The objective is to deconstruct the "black box" of AI by providing a deep dive into its core algorithms and models.
This tier begins with the foundational prerequisite for all natural language processing tasks: the digitization of text. It systematically explains the progression from one-hot encoding and the bag-of-words model to more sophisticated methods like TF-IDF and word embeddings. It also introduces the mathematical method of using Euclidean distance to calculate data similarity.
Building on this foundation, the series introduces the basic paradigms of machine learning. It covers supervised learning, using K-Nearest Neighbors and Decision Trees as examples, and unsupervised learning, demonstrated with the K-Means clustering algorithm. Students are guided through manual calculations to understand the procedural logic of these algorithms.
Neural networks constitute a central focus of this tier. The curriculum begins with an analogy to biological neurons and progressively explains the mechanics of network training, including weights, biases, forward propagation, and backpropagation. It then details specialized architectures such as Convolutional Neural Networks (CNNs) for image data and Recurrent Neural Networks (RNNs) for sequential data.
The content further advances to cover the Transformer architecture and its self-attention mechanism, which underpins modern large language models, as well as the game-theory principles of Generative Adversarial Networks (GANs).
In addition to models, this tier deepens practical skills and theoretical knowledge. It teaches advanced prompt engineering techniques, such as few-shot and chain-of-thought prompting, and provides a more detailed deconstruction of speech recognition systems into their constituent acoustic and language models.
At a higher level of abstraction, the series proposes a unifying framework for understanding AI systems, structured around three pillars: data, algorithms, and computing power. This framework is then used to explore advanced AI reasoning paradigms, including search strategies, game theory, and causal inference.
The pedagogical method shifts from experience to inquiry. Learning is facilitated through the deconstruction of algorithmic steps, formal reasoning, and complex thought experiments like the Prisoner's Dilemma to develop students' analytical and abstract thinking capabilities.
The "Comprehensive Applications" Volume – The Leap from Theory to Code
The "Comprehensive Applications" volume serves as the capstone of the series, primarily targeting students at the high school level and above. Its core objective is to facilitate the transition from theoretical understanding to practical implementation through code-based, comprehensive projects.
The most significant shift at this stage is the transition in tooling. Students move from graphical programming interfaces to an industry-standard development environment, utilizing the Python programming language and the PyTorch deep learning framework to complete all projects.
In the domain of computer vision, students are tasked with building a complete flower image classifier. This project encompasses the entire workflow, from dataset preparation and CNN model definition to hyperparameter tuning and performance evaluation.
For natural language processing, the central project is the construction of a movie review sentiment analyzer. Students learn to process text using word embedding techniques and to build and train an RNN model for the classification task.
The final module on generative AI combines a creative project—producing a multimedia story—with a structured ethical exercise. Students engage in a hands-on deepfake creation activity to directly observe its potential impact, followed by a systematic discussion of major challenges such as data privacy, algorithmic bias, and misinformation.
Summary
In summary, this textbook series establishes an integrated AI general education curriculum that spans from primary school to undergraduate studies, organized into three tiers—foundational, advanced, and application—that facilitate a spiral progression of knowledge and skills.
A defining feature of its educational path is the guided transition of the student's role, moving from a consumer of AI technology to a creator through a deliberate shift in pedagogical tools from graphical interfaces to code-based frameworks.
When compared to modular, activity-based resource libraries like MIT's Day of AI, which emphasize flexible and widespread engagement, this series represents a more systematic, long-term "curriculum" approach aimed at building a deep and coherent knowledge base.
This standardized, top-down curriculum design reflects an educational strategy focused on ensuring broad and equitable access to a consistent and high-quality learning pathway, with the goal of establishing a robust foundation for national AI literacy.